On Comparison of Classical and Machine Learning Models for Forecasting Rabi Green Gram Production in Odisha
241 / 105
Keywords:
: Akaike Information Criteria Corrected; Forecasting; Hybrid ARIMA-ANN; linearity; Non- linearity.Abstract
Green gram is one of the major pulse crop of Odisha grown mostly during the Rabi seasons. Rabi green gram production accounts for approximately
51 percent of the total pulse production in Odisha. Data on production (in ‘000 MT) of rabi green gram are collected for the period from 1970-71 to
2023-2024. The entire data is splitted into two sets: training data set and testing/Cross validated data set. The data for the period from 1970-71 to 2019
20 for building training data set and 2020-21 to 2023-24 to be used as test data set/ cross validation data set. Classical Auto Regressive Integrated
Moving Average (ARIMA) model and Artificial Neural Network (ANN) model is built for the training set data. Both the built-in models are compared
on basis of Mean Absolute Percentage Error MAPE and Akaike Information Criteria Corrected (AICc) for the training set and test set data. The model
performing better on basis of MAPE and AICc is selected for obtaining final forecast values of rabi green gram production.
ARIMA (2, 0, 0) fits best which after performing the BDS test for the residuals shows that both linearity and non-linearity is found in the data in ratio
of 7:1. So hybrid ARIMA-ANN model is prescribed to be fitted with appropriate weightage to selected best fit ARIMA and ANN models. Out of the
suitable NNAR models, NNAR (2, 3) model to be the best fit. So the hybrid ARIMA-ANN model is fitted with weightage of 0.875 to the selected
best fit ARIMA (2,0,0) model and 0.125 to the selected best fit NNAR (2,3) model. The model fit statistics for the hybrid ARIMA-ANN model is
found to be lower in case of training data set and testing data set. The forecast values of rabi green gram production obtained from the fitted hybrid
ARIMA-ANN model are found to be decreasing in future years. Thus, it could be concluded that the advanced machine learning model definitely
improves the accuracy of forecasting.
Downloads
References
1. Dash, A, Dhakre, DS and Bhattacharya D.(2017). Fitting of appropriate
2. model to study growth rate and instability of mango production in
3. India. Agric. Sci. Digest., 37(3) 2017: 191-196
4. Dash, A.Mangaraju, A, Mishra, P. and Nayak H. (2020). Using
5. Autoregressive Integrated Moving Average(ARIMA) Technique
6. to Forecast the Production of Kharif Cereals in Odisha (India).
7. Current Journal of Applied Science and Technology, 39(9):
8. 104-113
9. Dash, A.Dhakre, D.S. and Bhattacharya D. (2017). Study of Growth
10. and Instability in Food Grain Productionof Odisha: A Statistical
11. Modelling Approach, Environment & Ecology, 35(4D) : 3341-3351
12. Dash, A, Dhakre, D.S. and Bhattacharya, D. (2017). Analysis of
13. Growth and Instability in Pulse Production of Odisha during Rabi
14. Season: A Statistical Modelling Approach, International Journal
15. of Current Microbiology and Applied Science, 6(8): 107-115
16. TABLES AND FIGURES
17. 363
18. Dash, A. and Hansdah, R. (2020). Analytical study of growth and
19. instability of rabi oilseed production in Odisha. Journal of
20. Pharmacognosy and Phytochemistry 2020, 9(2), 2145-2149
21. Devi M, Kumar J, Malik D P and Mishra, P.(2021). Forecasting of
22. wheat production in Haryana using hybrid time series model.
23. Journal of Agriculture and Food Research, 5, 100175.
24. Dibya, B and Dash, A. (2022). Using ARIMA Model to Forecast the
25. Area, Yield and Production of Arhar in Odisha. Biological Forum.
26. 14(3), 1179-1185
27. Garai, S., Paul, R.K., Yeasin, M. and Paul, A.K. (2024). CEEMDAN
28. Based Hybrid Machine Learning Models for Time Series
29. Forecasting Using MARS Algorithm and PSO-Optimization.
30. Neural Processing Letters, https://doi.org/10.1007/s11063-024
31. 11552-w
32. Mishra, P. Khatib, AMGA, Lal P, Anwar, A.N., Korakot, Abotaleb,
33. M, Ray, S and Veerasak, P. (2023). An Overview of Pulses
34. Production in India: Retrospect and Prospects of the Future Food
35. with an Application of Hybrid Models. National Academy Science
36. Letters. 46, 366-374
37. Mitra, D. and Paul, R.K. (2017). Hybrid time-series models for
38. forecasting agricultural commodity prices. Model Assisted
39. Statistics and Applications, 12, 255-264.
40. Paul, R.K., Das, T. and Yeasin, M. (2023). Ensemble of Time Series and
41. Machine Learning Model for Forecasting Volatility in Agricultural
42. Prices. National Science Academy Science Letter.https://doi.
43. org/10.1007/s40009-023-01218-x
44. Pradhan, J and Dash, A. (2024). Using ARIMA Model to Forecast
45. Production of kharif Sweet Potato in Odisha.Environment and
46. Ecology,. 42(2):383-390
47. Raghav Y S, Mishra P, Alakkari K M, Singh M, Al Khatib A M G and
48. Balloo, R (2022). Modelling and forecasting of pulses production
49. in south Asian countries and its role in nutritional security. Legume
50. Research-An International Journal, 45(4), 454-461.
51. Rakshit, D., & Paul, R.K. (2024). Hybrid ARFIMA-LRNN Model for
52. Forecasting Commodity Prices. Journal of the Indian Society of
53. Agricultural Statistics, 78(3): 221-229
54. Sahu, G and Dash, A. (2022). Forecasting of Area, Yield and Production
55. of Mustard in Odisha. Biological Forum. 14(3): 1157-1163
56. Sameerbanu, P. (2024). Forecasting Onion Production in India Using
57. Arima Models: A Research Study. International Journal of
58. Scientific Research in Engineering and Management. 8(8): 1-7
59. Sankar, T.J. and Pushpa, P.(2025). Design and Development of ARIMA
60. Model for Allium Cepa Production in India. Palestine Journal of
61. Mathematics, 14.
62. Yeasin, M., Paul, R.K. & Shankar, S.V. (2024). Ensemble machine
63. learning models for forecasting tropical cyclones in North Indian
64. region. Earth Sci Inform. https://doi.org/10.1007/s12145-024
65. 01366-8